To sell data, a company licenses a copy of records it owns to a buyer under a written agreement: directly to an AI developer or other buyer, through a data marketplace, through a reseller, or as part of a wind-down or sale of the business. What sells is data buyers cannot collect themselves, with clear rights, years of history and personal information removed; public deals run from about $25,000 for a closed startup's code base to a $10 million winning bid for a bankrupt airline's emails and chats.
- Sell a license, not the records: your company keeps the data and grants a buyer defined rights to a copy.
- Marketplaces suit structured datasets you can refresh and sell by subscription. Internal records licensed once for AI training fit a direct license better.
- Check ownership, customer contracts and privacy promises before any record leaves. Quietly changing a privacy policy to allow a sale can be unfair or deceptive, the FTC has warned.
What kinds of company data sell?
Buyers pay for data they cannot scrape or generate themselves, and they pay more when rights are clear and documented, according to Origin's AI data licensing deals tracker. Sales of company records are usually private, so the public examples below show that a market exists rather than set a price.
For what each record type is worth to AI buyers, see how much AI companies pay for data.
| Data | Who buys it | Public example |
|---|---|---|
| Internal email, chat and documents | AI developers and technology companies | A bankrupt airline's archive of about 100 million emails and 500 million chats drew a $10 million winning bid (Aug 2026) |
| Workspace archives from closed startups: Slack, email, docs | AI developers and data vendors building simulated workplaces for AI agents | Some deals reportedly reached six figures |
| Source code | AI developers | About $25,000 for one closed startup's code base |
| Images, video and 3D scans | AI and robotics developers | An aerial survey firm licensed existing 3D and LiDAR data for A$1.07 million combined |
| Structured datasets: prices, transactions, locations, weather | Analysts, investment funds and data science teams | Sold by subscription on data marketplaces; one cloud data exchange lists more than 3,500 data products |
Source: Origin Data Partners: AI data licensing deals tracker, 2023 to 2026
What data is hard or illegal to sell?
Rule these out first. Each one can stop a sale on its own.
- Unanonymised personal data. One B2B data marketplace bans unanonymised sensitive personal data outright, and cloud marketplaces set similar rules (compared here).
- Data you do not own outright: records licensed from a software vendor, or held for clients under contract. Marketplaces require the legal right to commercialise what you list.
- Illegally scraped data. Marketplaces ban it.
- Thin data. Buyer programs set size minimums from 5 staff to 30 full-time employees, and one asks for three or more years of history, Origin's buyer programs study found.
- Bulk sensitive data such as health, biometric, genetic, financial or precise location records. A US Justice Department program in effect since April 8, 2025 restricts access to it by foreign adversaries.
Who buys data from companies?
Buyers fall into four broad groups, and each wants something different.
- AI developers: labs, AI data companies and builders of agent training environments. They license internal records such as tickets, email, documents and code to train models and build simulated workplaces for agents.
- Investment funds and the data scouts that serve them. They buy structured datasets that signal company performance, usually by subscription. See where to sell data to hedge funds.
- Analytics and data science teams that buy through cloud data marketplaces.
- Acquirers, liquidators and wind-down firms, which buy or sell data with the rest of a business.
Where can you sell data? Four ways compared
The four routes differ in who the buyer is, how much work you do and how you are paid. Origin's sell company data hub covers the direct route by country and industry.
| Route | Who buys | How it works | How you are paid | Best fit |
|---|---|---|---|---|
| Direct license to an AI developer | AI labs, AI data companies and agent-environment builders | You agree scope, permitted use, de-identification and price in a written license | A fee set in the license; published payment timing runs from 2 to 4 weeks to about 60 days | Internal operational records: tickets, email, documents, code |
| Data marketplace or exchange | Customers of that marketplace: analysts, data science teams, funds | You list a documented dataset with samples; buyers subscribe or request access | Subscription or usage fees, minus a listing fee or commission | Structured datasets you can refresh |
| Reseller or aggregator | The reseller's clients, such as investment firms | A partner packages your data and sells it, sometimes under its own brand | Set by the reseller | Owners who want no sales or delivery work |
| Wind-down or sale of the business | AI developers, through wind-down firms, auctions or the acquirer | Records are sold or licensed with other assets; a court or assignee may run the sale | A one-time sale price | Companies that are closing or being sold |
How each route works
Direct license. You describe the records, the buyer reviews a sample and proposes terms, and you deliver a de-identified copy. Origin's study of 8 programs that pay companies for internal records found published payment timing from 2 to 4 weeks to about 60 days, and one program values records within about a week of sampling.
Marketplace. You build a listing with documentation and samples, and the marketplace handles discovery and often billing. Marketplaces expect regular updates and subscriber support, and all set strict rules on personal data. See data marketplaces compared for each one's fees and rules.
Reseller. A partner takes on packaging, compliance, contracts and delivery. One financial-data reseller sells data under its own brand to hedge funds, asset managers, investment banks and private equity, and says first payments can come within six months.
Wind-down or sale. Data can be sold with the business or on its own when a company closes. Read licensing data when selling a business and what has value after closing.
Source: Origin Data Partners: what AI buyers publish for company data, 8 programs compared (Oct 2026)
How is data priced?
There is no public rate card for company data. Price depends on the records, the years they cover, the rights you hold and what the buyer may do with them. Deals use four structures.
| Structure | How it works | Public example |
|---|---|---|
| One-time fee | A single payment for a defined copy and use | An academic publisher's $23M one-time deal; about $25,000 for a closed startup's code base |
| Recurring license | Annual or multi-year payments for continuing access | About $60M a year for a platform's posts and comments; more than $250M over five years for a news publisher's archive and reporting |
| Per record or per item | A price per photo, title, hour or file | Asking prices of roughly 5 cents to $1 per photo; $5,000 per book title for three years |
| Subscription or usage fee | Marketplace buyers pay monthly, yearly or per query | See data marketplaces compared |
Source: Origin Data Partners: AI data licensing deals tracker, 2023 to 2026
What do buyers publish for company records?
For an ordinary company, the best guide is what buyer programs publish. In Origin's buyer programs study, the typical published range for a company was about $50K at the low end and $843K at the top, and the top of a range was a median 12 times the bottom. Where a company lands depends far more on its records than on its industry.
These are ceilings set by each program, not offers. One program advertises $100K to $4M+ while its own calculator shows up to $100,000. Ask for a written offer before comparing.
Terms matter as well as the headline. One intermediary in the study takes perpetual exclusivity by default; exclusivity should be priced separately. See exclusive vs non-exclusive licenses and data licensing pricing models.
Source: Origin Data Partners: what AI buyers publish for company data, 8 programs compared (Oct 2026)
What should you check before selling data?
Answer these before you contact a buyer. This is not legal advice; ask counsel to review the specific records and contracts.
| Check | What to look at | Why it matters |
|---|---|---|
| Ownership | Did your company create the records, or license them from a vendor or client? | Marketplaces require the legal right to commercialise what you list, and buyers check it |
| Customer and vendor contracts | Confidentiality clauses, data processing agreements and limits on reuse | Records about clients can be restricted even though your company holds them |
| Privacy promises | Your privacy policy and what staff and customers were told when the data was collected | The FTC has warned that quietly changing terms to allow sharing or AI training may be unfair or deceptive, and ordered a software company to pay $16.5 million after it sold browsing data despite privacy promises |
| State privacy law | Personal information of California residents and others covered by state law | Under the CCPA, consumers can tell businesses to stop selling or sharing their personal information, and businesses that sell it must offer a "Do Not Sell or Share My Personal Information" link |
| Sensitive data | Health, biometric, genetic, financial and precise location data | The Justice Department's Data Security Program restricts access to bulk sensitive personal data by foreign adversaries |
| Bankruptcy | Whether the company is in bankruptcy and promised not to transfer personal data | Section 363 allows that data to be sold only consistent with the policy or after a consumer privacy ombudsman is appointed and the court approves |
Source: FTC: quietly changing your terms of service could be unfair or deceptive (Feb 2024)
How do companies sell data without breaking privacy promises?
Remove or replace personal information before delivery, and keep the license narrow. In a retailer's bankruptcy, the FTC recommended that customer data go only to a buyer in substantially the same line of business that agreed to be bound by the original privacy policies.
For the details, read is it legal to sell company data for AI training, how to de-identify company data and GDPR, CCPA and licensing company data.
Source: FTC: conditions on selling a retailer's customer data in bankruptcy (May 2015)
How to sell data, step by step
Each step is a separate decision, and you can stop at any of them.
- List what you hold at company level: systems, record types, years covered and volume. No export is needed yet. The data inventory template gives you the columns.
- Run the checks above for each record type: ownership, contracts, privacy promises and sensitive data.
- Pick the route: a direct license for internal records, a marketplace for a dataset you can refresh, a reseller if you want no sales work, or an asset sale if the company is closing.
- Prepare a one-page description and a small de-identified sample to share once a buyer is interested.
- Compare terms, not just headline price: permitted uses, exclusivity, term, deletion and payment timing. Use the due-diligence checklist and what a data licensing agreement should include.
- Deliver the agreed copy, de-identified as the license requires, and keep a record of exactly what was sent.
- Get paid on the agreed schedule. Published timing runs from 2 to 4 weeks to about 60 days.
What are the red flags when selling data?
- A buyer wants a full export before terms are signed. Share a description and a sample first.
- A headline payout with no written offer. Published ranges are ceilings, and headlines can run far above a program's own calculator.
- Perpetual exclusivity by default, with no extra payment for it.
- No written limit on use. The license should say whether the buyer may train models, resell the data or share it.
- A plan to change your privacy policy just before the sale. The FTC has warned that this may be unfair or deceptive.
- Pressure to include customer personal information the buyer does not need.
Next step
Pick one record type with outcomes attached, such as support tickets or work orders, and describe it at company level: the system, years covered and volume. Then compare what a direct license, a marketplace and a reseller would mean for that record type. See who buys company data for AI training for what each buyer type wants.
Origin Data Partners is an independent data partner that helps companies get their data bought. It is free to the company that owns the data; a buyer may pay Origin when a deal closes. See what your data is worth.
Common questions
How do I sell my company's data?
List what you hold, check that you own it and that contracts and privacy promises allow a sale, remove personal information, then license a copy to a buyer under a written agreement. The main routes are a direct license to an AI developer, a data marketplace, a reseller, or a sale with the business when it closes.
Where can I sell data?
Directly to AI developers and other buyers under a license, on data marketplaces and cloud data exchanges, through resellers that package and sell data for you, or as part of a wind-down or sale of the business. Marketplaces suit structured datasets you can refresh; internal records licensed once for AI training fit a direct license better.
Who buys data?
Four broad groups: AI developers that license internal records for training and agent environments, investment funds that subscribe to structured datasets, analytics teams that buy through cloud marketplaces, and acquirers or wind-down firms that buy data with a business.
How much can a company get for its data?
It depends on the records, their history and your rights. In Origin's study of buyer programs, the typical published range was about $50K to $843K, and these are ceilings, not offers. Public deals include about $25,000 for a closed startup's code base and a $10 million winning bid for a bankrupt airline's emails and chats.
Can I sell a database?
Yes, if your company owns it and the people in it were not promised it would stay private. Databases with customer personal information face extra rules: marketplaces ban unanonymised sensitive personal data, and state privacy laws such as the CCPA give consumers the right to opt out of the sale of their data.
Is it legal to sell customer data?
Sometimes, but only within what customers were told and what privacy law allows. The FTC has warned that quietly changing terms to allow sharing or AI training may be unfair or deceptive, and California consumers can opt out of sales. De-identifying the data first and licensing it narrowly lowers the risk.
Do I give up my data when I sell it?
Not if you license it. A license lets your company keep the records and grant the buyer defined rights to a copy, for a set use and term. Exclusivity, if the buyer wants it, should be priced separately.
Sources and further reading
Origin Data Partners is independent and not affiliated with any buyer named on this page.
- Origin Data Partners: AI data licensing deals tracker, 2023 to 2026
- Origin Data Partners: what AI buyers publish for company data, 8 programs compared (Oct 2026)
- FTC: quietly changing your terms of service could be unfair or deceptive (Feb 2024)
- FTC: order bans a software company from selling browsing data, $16.5 million (Feb 2024)
- FTC: conditions on selling a retailer's customer data in bankruptcy (May 2015)
- California Attorney General: California Consumer Privacy Act
- 11 U.S. Code section 363 (sale of property in bankruptcy)
- US Justice Department: Data Security Program on Americans' sensitive data (Apr 2025)
- A B2B data marketplace's provider rules
- A cloud data exchange's FAQs
- A financial-data reseller's go-to-market service for data owners